C H A P T E R

1
A Systematic View of Remote Sensing
O U T L I N E

1.1. Introduction 2 1.6.1.1. Scene Generation 22

1.6.1.2. Surface Radiation
1.2. Platform and Sensor System 2
Modeling 22
1.2.1. Geostationary Satellites 3
1.6.1.3. Atmospheric Radiative
1.2.2. Polar-orbiting Satellites 4
Transfer 24
1.2.3. U.S. Operational Missions 5
1.6.1.4. Sensor Modeling 25
1.2.4. Sensor Types 7
1.2.5. Data Characteristics 8
1.6.2. Inversion Methods 25
1.2.5.1. Spatial Resolution 8
1.6.2.1. Statistical Analysis and
1.2.5.2. Spectral Resolution 8
Machine Learning
1.2.5.3. Temporal Resolution 8
Techniques 25
1.2.5.4. Radiometric Resolution 8
1.6.2.2. Optimization Algorithms 26
1.3. Data Transmission and Ground Receiving 1.6.2.3. Look-up Table
System 11 Algorithms 26
1.6.2.4. Data Assimilation
1.4. Data Processing System 14
Methods 27
1.4.1. Radiometric Calibration 14
1.4.2. Geometric Processing 15 1.7. Production, Archiving, and Distribution
1.4.3. Image Quality Enhancement 16 of High-level Products 27
1.4.4. Atmospheric Correction 17
1.8. Product Validation 28
1.4.5. Image Fusion and Product
Integration 18 1.9. Remote Sensing Applications 28
1.5. Mapping Category Variables 18 1.10. Concluding Remarks 30
1.6. Estimating Quantitative Variables 22
1.6.1. Forward Radiation Modeling 22

Advanced Remote Sensing DOI: 10.1016/B978-0-12-385954-9.00001-0 1 Copyright Ó 2012 Elsevier Inc. All rights reserved.
2 1. A SYSTEMATIC VIEW OF REMOTE SENSING

Abstract
This chapter provides an overview of the remote-sensing
system, including the platform and sensor system, data
transmission and ground receiving system, processing
system of radiometric and geometric properties, analysis
system for mapping category variables and generating
high-level products of quantitative variables, product
production and distribution system, product validation
system, and remote- sensing applications. It aims to present
a complete picture of the state-of-the-art development of
remote sensing techniques by linking different chapters in
the rest of the book and filling in any possible gaps.

1.1. INTRODUCTION
We are living in a world whose population is
rapidly increasing, depleting natural resources,
and experiencing the possible consequences of
human-induced climate change. Our ability to
meet these challenges partially depends on
how well we understand the Earth system and
use that information to guide our actions.
Remote sensing is a tremendous source of infor-
mation needed by policy makers, resource
managers, forecasters, and other users, and it
has become increasingly vital for the effective
and sustainable future management of the Earth.
A remote sensing system consists of instru- FIGURE 1.1 Key components of the remote sensing
mentation, processing, and analysis designed system.
to measure, monitor, and predict the physical,
chemical, and biological aspects of the Earth numerical variables of the Earth surface environ-
system. Sophisticated new technologies have ment, the product generation and distribution
been developed to gather vast quantities of system, the product validation system, and
data, and the mathematical and physical sophis- end-user applications. Applications largely
tication of the techniques used to process and define the data acquisition system, and end-
analyze the observed data has increased users often need to validate the products to
considerably. quantify their errors and uncertainties.
The first chapter of the book aims to link
diverse components to paint a full picture of
a remote sensing system as illustrated in 1.2. PLATFORM AND SENSOR
Figure 1.1. It starts with a brief introduction to SYSTEM
the platform and sensor system for acquiring
data, and then moves on to the data transmission The data acquisition system mainly consists of
and ground receiving system, the processing the sensor and the platform on which the sensor
system for handling the geometric and radio- resides. The platform may be on the surface, in
metric properties of data, the analysis system the air, or in space. A surface platform may
for extracting information on both category and be a ladder, tower, cherry-picker, crane, tall
 1.2. PLATFORM AND SENSOR SYSTEM 3
building, or scaffolding that provides data used very close to 35,786 km (22,236 m) and keeps
primarily for validation. Aerial platforms the satellite fixed over one longitude at the
include aircraft and balloons. Unmanned aerial equator. The satellite appears motionless at
vehicles (UAVs) are being increasingly used for a fixed position in the sky to ground observers.
remote sensing purposes these days. Spaceborne There are several hundred communication satel-
platforms are mainly satellites and space shut- lites and several meteorological satellites in such
tles. Only satellite sensor systems (both geosta- an orbit. Figure 1.2 illustrates a few typical mete-
tionary and polar-orbiting) will be considered orological satellites in the geostationary orbit
in the rest of this section. relative to the polar-orbiting satellites.
U.S. operational weather satellites include the
geostationary operational environmental satellites
1.2.1. Geostationary Satellites
(GOES) for short-range warning and “now-
A geostationary satellite is in a geostationary casting” primarily to support the National
orbit, which can only be achieved at an altitude Weather Service (NWS) requirements. The

FIGURE 1.2 Illustration of the distribution of a few common geostationary satellites compared to the polar-orbiting
satellites.
4 1. A SYSTEMATIC VIEW OF REMOTE SENSING

procurement, design, and manufacturing of GOES named Fengyun (FY-2) from FY-2A to FY-2F
are overseen by the National Aeronautics and since 1997. The second generation of geostation-
Space Administration (NASA), while all opera- ary meteorological satellites FY-4 is also under
tions of the satellites once in orbit are effected by development.
the National Oceanic and Atmospheric Adminis-
tration (NOAA). Before being launched, GOES
satellites are designated by letters (-A, -B, -C.).
1.2.2. Polar-orbiting Satellites
Once a GOES satellite is launched successfully, it Polar-orbiting satellites can provide an obser-
is re-designated with a number (-1, -2, -3.). vational platform for the entire planet surface,
Normally two GOES satellites are operational. while their geostationary counterparts are
Information on the GOES series is shown in Table limited to approximately 60 degrees of latitude
1.1. The third generation of GOES, the new GOES-
R satellite series planned for launch in 2015, repre-
TABLE 1.1 Information on GOES Satellite Series
sents a significant improvement in spatial,
temporal, and spectral observations (several Satellites Launch day Status
orders of magnitude) over the capabilities of the
1 Oct. 16, 1975 decommissioned
currently operational GOES series.
European operational missions are currently 2 June 16, 1977 decommissioned
operated by the European Organisation for the 3 June 16, 1978 decommissioned
Exploitation of Meteorological Satellites
4 Sept. 9, 1978 decommissioned
(EUMETSAT). EUMETSAT’s geostationary
satellite programs include the earlier Meteosat 5 May 22, 1981 deactivated on July 18,
1990
system (up to Meteosat-7) and three Meteosat
Second Generation (MSG) satellites (MSG-1,2,3 6 April 28, 1983 decommissioned
or Meteosat-8,9,10). A fourth MSG satellite is G May 3, 1986 Failed to orbit
being considered. The MSG satellites carry an
7 April 28, 1987 used as
impressive pair of instrumentsdthe Spinning
a communications
Enhanced Visible and InfraRed Imager (SEVIRI), satellite
which has the capacity to observe the Earth in 12
8 April 13, 1994 decommissioned
spectral channels and provide image data every
half hour, and the Geostationary Earth Radiation 9 May 23, 1995 decommissioned
Budget (GERB) instrument supporting climate 10 April 25, 1997 decommissioned
studies. The Meteosat Third Generation is also
11 May 3, 2000 In operation as GOES-
in the planning stage and is expected to achieve
West
operation in the 2020s.
The Japanese Geostationary Meteorological 12 July 23, 2001 standby, providing
coverage for South
Satellite (GMS) series had five satellites from
America
1977. The Multifunctional Transport Satellites
(MTSAT) are the successors to the GMS 1-5 satel- 13 May 24, 2006 In operation as GOES
East
lite series. The launch of MTSAT-1 failed in 1999,
but MTSAT-1R was successfully launched in 14 June 27, 2009 On-orbit storage; Will
2005. After a five-year service in space, it was replace GOES-11
replaced by MTSAT-2 from 2010. 15 March 4, 2010 stand-by
China has launched six of the first generation
GOES-R Scheduled in 2015
of geostationary meteorological satellites
 1.2. PLATFORM AND SENSOR SYSTEM 5
at a fixed point over the earth. Polar-orbiting 21, 2000, which includes three advanced
satellites are able to circle the globe approxi- land imaging sensors and five revolutionary
mately once every 100 minutes. Most polar- cross cutting spacecraft technologies.
orbiting Earth observation satellites, such as The three sensors led to a new generation of
Terra, ENVISAT, and Landsat, have an altitude lighter weight, higher performance, and
of about 800 km. They are in sun-synchronous lower cost Landsat-type Earth surface
orbits passing directly over a given spot on the imaging instruments. The hyperspectral
ground at the same local time. A relatively low sensor Hyperion is the first of its kind to
orbit allows detection and collection of data by provide images of land surface in more than
instruments aboard a polar-orbiting satellite at 220 spectral bands.
a higher spatial resolution than from a geosta- Systematic missions provide systematic
tionary satellite. measurements of key environmental variables
NASA has launched a series of polar-orbiting that are essential to specify changes in
satellite missions with the ability to characterize forcings caused by factors outside the Earth
the current state of the Earth system. All the system (e.g., changes in incident solar
missions fall into three types: exploratory, oper- radiation) and to document the behavior of
ational precursor and technology demonstra- the major components of the Earth system. An
tion, and systematic. example is the Earth Observing System (EOS)
program. EOS is the centerpiece of NASA’s
Exploratory missions are to yield new
recent Earth observation program. It was
scientific breakthroughs. Each exploratory
conceived in the 1980s and began to take
satellite project is expected to be a one-time
shape in the early 1990s. It is composed of
mission that can deliver conclusive scientific
a series of satellites and sensors, a science
results addressing a focused set of scientific
component, and a data system supporting
questions. In some cases, an exploratory
a coordinated series of polar-orbiting and low
mission may focus on a single pioneering
inclination satellites for long-term global
measurement that opens a new window on
observations of the land surface, biosphere,
the behavior of the Earth system. These
solid Earth, atmosphere, and oceans.
missions are managed in the NASA Earth
Complete and still active EOS satellites are
System Science program (ESSP). Examples
shown in Tables 1.2 and 1.3.
include the Gravity Recovery and Climate
Experiment (GRACE), and CloudSAT.
GRACE data can be used for estimating soil
1.2.3. U.S. Operational Missions
moisture and surface/underground water
(Section 21.3). NOAA’s operational environmental satellite
Operational precursor and technology system is composed of two types of satellites:
demonstration missions enable major GOES for national, regional, short-range
upgrades of existing operational observing warning and “now-casting,” and polar-orbiting
systems. NASA is investing in innovative environmental satellites (POES) for global,
sensor technologies and developing more long-term forecasting, and environmental moni-
cost-effective versions of its pioneer scientific toring. Both types of satellite are necessary for
instruments that can be used effectively by providing a complete global weather monitoring
operational agencies. An example is the NMP system. In addition, NOAA operates satellites in
EO-1 (New Millennium Program Earth the Defense Meteorological Satellite Program
Observing-1) mission launched on November (DMSP) that are also polar-orbiting satellites.
6 1. A SYSTEMATIC VIEW OF REMOTE SENSING

TABLE 1.2 Active EOS Satellites as of November 2011 TABLE 1.3 Completed EOS Satellites as of November
2011
Satellites Launch day
Satellites Lunch day
Landsat-7 April 15, 1999
Upper Atmosphere Research Satellite September 12, 1991
Quik Scatterometer June 19, 1999 (UARS)
(QuikSCAT)
ATLAS March 24, 1992
Terra December 18, 1999
TOPEX/Poseidon August 10, 1992
Active Cavity Radiometer December 20, 1999
Irradiance Monitor Satellite Spaceborne Imaging Radar-C (SIR-C) April 19, 1994
(ACRIMSAT)
Radar Satellite (RADARSAT) November 4, 1995
Jason-1 December 7, 2001
Total Ozone Mapping Spectrometer- July 2, 1996
Aqua May 4, 2002 Earth Probe (TOMS-EP)

Solar Radiation and January 25, 2003 Advanced Earth Observing Satellite August 17, 1996
Climate Experiment (ADEOS)
(SORCE)
Orbview-2/SeaWiFS August 1, 1997
Aura July 15, 2004
Tomographic Experiment using May 18, 1999
Ocean Surface Topography June 20, 2008 Radiative Recombinative Ionospheric
Mission (OSTM) EUV and Radio Sources (TERRIERS)
Landsat Data Continuity December, 2012 Shuttle Radar Topography Mission February 11, 2000
Mission (LDCM) (SRTM)
Challenging Mini-Satellite Payload July 15, 2000
(CHAMP)
The GOES satellites have been introduced in Stratospheric Aerosol and Gas December 10, 2001
Section 1.2.1. The POES system includes the Experiment (SAGE III)
Advanced Very High Resolution Radiometer
SeaWinds (ADEOS II) December 14, 2002
(AVHRR) and the Television Infrared Observa-
tion Satellite (TIROS) Operational Vertical Ice, Cloud, and Land Elevation January 12, 2003
Satellite (ICESat)
Sounder (TOVS). The world’s first meteorolog-
ical satellite, TIROS, was launched on April 1,
1960, and demonstrated the advantage of
mapping Earth’s cloud cover from satellite ATN) satellites, designated NOAA-8, was
altitudes. launched. NOAA continues to operate the ATN
On January 23, 1970, the first of the improved series of satellites today with improved instru-
TIROS Operational Satellite (ITOS) was ments. Complementing the geostationary satel-
launched. Between December 11, 1970, and lites are two NOAA polar-orbiting satellites,
July 29, 1976, five ITOS satellites designated one crossing the equator at 7:30 AM local time,
NOAA-1 through 5 were launched. From the other at 1:40 PM local time. The latest is
October 13, 1978, to July 23, 1981, satellites in NOAA-19, launched on February 6, 2009.
the TIROS-N series were launched, where N NOAA-18 (PM secondary), NOAA-17 (AM
represents the next generation of operational backup), NOAA-16 (PM secondary), and
satellites. NOAA-6 and NOAA-7 were also NOAA-15 (AM secondary) all continue transmit-
launched during this time frame. On March 28, ting data as stand-by satellites. NOAA-19 is
1983, the first of the Advanced TIROS-N (or the “operational” PM primary satellite, and
 1.2. PLATFORM AND SENSOR SYSTEM 7
METOP-A, owned and operated by EUMET- scanning from one side of the sensor to the
SAT, is the AM Primary satellite. other across the platform flight direction
The first AVHRR sensor was a 4-channel radi- using a rotating mirror, are called Whiskbroom
ometer, first carried on TIROS-N (launched Scanners, such as AVHRR. Alone-track
October 1978). This was subsequently improved scanners, scanning a swath with a linear array
to a 5-channel instrument (AVHRR/2) that was of Charge Coupled Devices (CCD) arranged
initially carried on NOAA-7 (launched June perpendicular to the flight direction of the
1981). The latest instrument version is AVHRR/3, platform without using a mechanical rotation
with 6 channels, first carried on NOAA-15, device, are called Pushbroom scanners, such as
launched in May 1998. Multiple global vegetation High Resolution Visible of SPOT and
index datasets have been developed from Advanced Land Imager of EO-1.
NOAA-7 to now. • Spectroradiometer: A radiometer that can
From 2011, NOAA has started the new “Joint measure the radiance in multiple spectral
Polar Satellite System” (JPSS) program that will bands, such as the Moderate Resolution
consist of platforms based on the NPOESS Prepa- Imaging Spectroradiometer (MODIS) and the
ratory Project (NPP) satellite, launched in October Multi-angle Imaging SpectroRadiometer
2011. NOAA and EUMETSAT have also estab- (MISR).
lished the Initial Joint Polar-Orbiting Operational
Active sensors provide their own electromag-
Satellite System (IJPS), which comprises two
netic radiation to illuminate the scene they
polar-orbiting satellite systems and their respec-
observe. They send a pulse of energy from the
tive ground segments from each side. It provides
sensor to the scene and then receive the radiation
meteorological data for “Morning” and “After-
that is reflected or backscattered from that scene.
noon” orbits by complementing each other’s
Typical active sensors include:
polar satellite global coverage.
• Radar (Radio Detection and Ranging): A
1.2.4. Sensor Types microwave radar that uses a transmitter
operating at microwave frequencies to emit
There are two types of sensorsdpassive and
electromagnetic radiation and a directional
active. Passive sensors detect natural radiation
antenna or receiver to measure the time of
that is emitted by the object being viewed or
arrival of reflected or backscattered pulses of
reflected by the object from a source other than
radiation from distant objects for determining
the instrument. Reflected sunlight is the most
the distance to the object.
common external source of radiation sensed by
• Synthetic-aperture radar (SAR): A side-looking
passive sensors. Typical passive sensors include:
radar imaging system that uses relative
• Radiometer: An instrument that quantitatively motion between an antenna and the Earth
measures the radiance of electromagnetic surface to synthesize a very long antenna by
radiation in the visible, infrared, or combining signals (echoes) received by the
microwave spectral region. radar as it moves along its flight track for
• Imaging Radiometer: A radiometer that obtaining high spatial resolution imagery.
includes a scanning capability to provide There are multiple SAR systems in operation,
a two-dimensional array of pixels from which and some examples can be seen in Section
an image may be produced. It is often called 19.3.2 for mapping soil moisture.
a scanner. Scanning can be performed • Interferometric synthetic aperture radar (InSAR):
mechanically or electronically by using an A technique that compares two or more
array of detectors. Across-track scanners, amplitude and phase images over the same
8 1. A SYSTEMATIC VIEW OF REMOTE SENSING

geographic region received during different 1.2.5. Data Characteristics

passes of the SAR platform at different times.
InSAR can survey height information of the The specifications of the platform and the
illuminated scene with cm-scale vertical sensor determine the resolutions of the remotely
resolution and 30-m pixel resolution, and sensed data: spatial, spectral, temporal, and
covering areas 100 km x 100 km (in standard radiometric.
beam modes). Examples include ERS-1 (1991), 1.2.5.1. Spatial Resolution
JERS-1 (1992), RADARSAT-1 and ERS-2
(1995), and ASAR (2002). While the majority Spatial resolution is a measure of the smallest
of InSAR missions to date have utilized object that can be resolved by the sensor, or the
C-band sensors, recent missions such as ground area imaged for the instantaneous field
ALOS PALSAR, TerraSAR-X, and COSMO of view (IFOV) of the sensor, or the linear dimen-
SKYMED are expanding the available data sion on the ground represented by each pixel.
in the L- and X-bands. Figure 1.3 shows the campus of the University
• Scatterometer: A high frequency microwave of Maryland at College Park at four different
radar designed specifically to determine the spatial resolutions. Table 1.4 shows the spatial
normalized radar cross section of the surface. resolution of some common sensors.
Over ocean surfaces, measurements of
backscattered radiation in the microwave 1.2.5.2. Spectral Resolution
spectral region can be used to derive maps of The spectral resolution describes the number
surface wind speed and direction. It has also and width of spectral bands in a sensor system.
been used for mapping surface soil moisture Many sensor systems have a panchromatic
and freeze/thaw states. Examples include the band, which is one single wide band in the
Advanced Microwave Instrument (AMI) of visible spectrum, and multispectral bands in
ERS-1 and ERS-2. the visible-near-IR or thermal-IR spectrum (see
• Lidar (Light Detection and Ranging): An active Table 1.4). Hyperspectral systems usually have
optical sensor that uses a laser in the hundreds of spectral narrow bands; for example,
ultraviolet, visible, or near infrared spectrum Hyperion on EO-1 satellite has 220 bands at 30-m
to transmit a light pulse and a receiver with spatial resolution.
sensitive detectors to measure the
backscattered or reflected light. Distance to 1.2.5.3. Temporal Resolution
the object is determined by recording the time Temporal resolution is a measure of the repeat
between the transmitted and backscattered cycle or frequency with which a sensor revisits
pulses and using the speed of light to calculate the same part of the Earth’s surface. The
the distance traveled. The details are given in frequency characteristics are determined by the
Chapters 14 and 15. design of the satellite sensor and its orbit pattern.
• Laser Altimeter: A laser altimeter that uses The temporal resolutions of common sensors are
a lidar to measure the height of the instrument also shown in Table 1.4.
platform above the surface. By independently
knowing the height of the platform with 1.2.5.4. Radiometric Resolution
respect to the mean Earth’s surface, the Radiometric resolution refers to the dynamic
topography of the underlying surface can be range, or the number of different output numbers
determined. The Geoscience Laser Altimeter in each band of data, and is determined by the
System (GLAS) of ICESat is a typical example number of bits into which the recorded radiation
of a space-based Laser Altimeter. is divided. In 8-bit data, the digital numbers (DN)
 1.2. PLATFORM AND SENSOR SYSTEM 9

(a) 1m

(b) 10m

FIGURE 1.3 Campus of the University of Maryland at College Park at four spatial resolutions.
10 1. A SYSTEMATIC VIEW OF REMOTE SENSING

(c) 30m

(d) 250m

FIGURE 1.3 (Continued).

 1.3. DATA TRANSMISSION AND GROUND RECEIVING SYSTEM 11
TABLE 1.4 The basic characteristics of some commonly used sensors

Spatial Radiometric Temporal Temporal

Satellite sensors Spectral bands resolution (m) resolution (bits) resolution (day) coverage
Coarse POLDER B1-B9 6*7 12 4 POLDER 1:
resolution October 1996 to
(> 1000 m) June 1997
POLDER2: April
to October 2003
Medium MODIS B1-b2 250 12 daily 1999
resolution
(100e1000 m) B3-b7 500
B8-b36 1000
AVHRR B1-B5 1100 at nadir 10 daily

Fine resolution ALI/EO1

(5-100 m)
ASTER/Terra B1 15 8
B2-B9 30
B11-B14 90 12
ETMþ/ Pan 15 8 16 1999-
Landsat7
B1-B5,B7 30

B6 60
HRV/SPOT5 Pan 2.5 or 5 8 26/2.4 2002-
B1-B3 10
SW-IR 20
Very high Ikonos Panchromatic 0.82 at nadir 11 3 days at 40 1999-
resolution band latitude
(< 5 m)
B1-b4 3.2 at nadir
Quickbird Pan 0.61 11 1-3.5 2001-

B1-B4 2.44
World view pan 0.5 at nadir 11 1.7-5.9 2007-
Geoeye-1 pan 1.41 at nadir 11 2.1-8.3 days at 2008-
40 latitude
B1-B4 1.65 at nadir

can range from 0 to 255 for each pixel (28 ¼ 256 total 1.3. DATA TRANSMISSION AND
possible numbers). Obviously more bits results GROUND RECEIVING SYSTEM
in higher radiometric accuracy of the sensor, as
shown in Figure 1.4. The radiometric resolutions There are three main options for transmit-
of common sensors are shown in Table 1.4. ting data acquired by satellite sensors to the
12 1. A SYSTEMATIC VIEW OF REMOTE SENSING

(a) 8 bits (256 levels)

(b) 4 bits (16 levels)

FIGURE 1.4 Four radiometric resolutions.

 1.3. DATA TRANSMISSION AND GROUND RECEIVING SYSTEM 13

(c) 2 bits (4 levels)

(d) 1 bit (2 levels)

FIGURE 1.4 (Continued).

14 1. A SYSTEMATIC VIEW OF REMOTE SENSING

surface: (1) the data can be directly transmitted means to perform acquisition in a remote loca-
to Earth if a Ground Receiving Station (GRS) is tion for a long period of time when a lot of
in the line of sight of the satellite; (2) the data images are needed for particular work (cartog-
can be recorded on board the satellite for trans- raphy of a region for example).
mission to a GRS at a later time; and (3) the The ground receiving stations acquire,
data can also be relayed to the GRS through preprocess, archive, and process data. Their
the Tracking and Data Relay Satellite System typical components and functions may include
(TDRSS), which consists of a series of commu- the data acquisition facility, the data processing
nications satellites in geosynchronous orbit. facility, the value added facility, and user
The data are transmitted from one satellite to support services.
another until they reach the appropriate GRS.
There are two types of GRSs: fixed and
mobile. Most GRSs are fixed, and Figure 1.5 1.4. DATA PROCESSING SYSTEM
shows the locations of all ground stations oper-
ated by the U.S. and International Cooperator Two types of preprocesses are conducted:
ground station network for the direct downlink radiometric processing and geometric processing.
and distribution of Landsat 7 (L7) and/or
Landsat 5 (L5) image data.
1.4.1. Radiometric Calibration
Since coverage of the globe by ground
receiving stations is not complete, as seen from Radiometric calibration is a process that
Figure 1.5, the mobile station is an attractive converts recorded sensor voltages or digital
solution to fill the holes and also an efficient numbers (DN) to an absolute scale of radiance

FIGURE 1.5 Landsat ground stations and their coverage. The circles show the approximate area over which each station has
the capability for direct reception of Landsat data. The green circles show the components of the L7 ground station network,
the red circles show components of the L5 station network, and the dashed circles show stations with dual (L5 and L7). status.
The yellow circles show L5 short-term (“campaign”) stations that may contribute to the USGS Landsat archive. (http://
landsat.usgs.gov/about_ground_stations.php)
 1.4. DATA PROCESSING SYSTEM 15
or reflectance. Because outer space is such a harsh vegetation, and relatively high surface reflectiv-
environment, the performance of all satellite ity with approximately Lambertian reflectance.
sensors degrades over time. To achieve consis- Commonly used sites include stable desert areas
tent and accurate measurements that can be of the Sahara and Saudi Arabia, the Sonoran
used to detect climatic and environmental Desert, the White Sand, and regions in Bolivia.
change, the DNs need to be transformed into
physical quantities.
Calibration measurements can be conducted
1.4.2. Geometric Processing
in three stages. Pre-flight calibration measures No image acquired by sensors can perfectly
a sensor’s radiometric properties before that represent the true spatial properties of the land-
sensor is sent into space. Pre-flight instrument scape. Many factors can also distort the geometric
calibration is performed at the instrument properties of remote sensing data, such as varia-
builder’s facilities. The controllable and stable tions in the platform altitude, attitude and
environment in the laboratory guarantees high velocity, Earth rotation and curvature, surface
calibration accuracy and precision. relief displacement, and perspective projection.
In-flight calibration is usually performed on Some of these resulting distortions are systematic
a routine basis with on-board calibration and can be corrected through analysis of sensor
systems. More and more optical sensors have characteristics and platform ephemeris data, but
on-board calibration devices. For example, the others are random and have to be corrected by
AVHRR optical sensor does not have an on- using ground control points (DCP).
board calibration capability, but the ETMþ has In the sensor ground instantaneous field of
three on-board calibration devices: the Internal view (IFOV), surface elements do not contribute
Calibrator, the Partial Aperture Solar Calibrator, to the pixel value equally, but rather, the central
and the Full Aperture Solar Calibrator. MODIS part contributes most to the pixel value. This
also has three dedicated calibration devices for kind of spatial effect is usually specified by the
the reflective bands: Solar Diffuser, Solar sensor point spread function (PSF) in the spatial
Diffuser Stability Monitor, and the Spectroradio- domain, and the Fourier transform of the PSF is
metric Calibration Assembly. In addition, called the modulation transfer function (MTF)d
MODIS has two additional calibration tech- a precise measurement of details and contrast
niques: looking at the Moon and at deep space. made in the frequency domain. The sensor PSF
Post-launch calibration data have to be obtained is often modeled as a Gaussian. Figure 1.6 illus-
from vicarious calibration techniques that typi- trates the PSF in two- and three-dimensions,
cally make use of selected natural or artificial sites where TGSD is the threshold ground sample
on the surface of the Earth. Pre-launch and distance, which is the centroid-to-centroid
onboard methods are better established, and the distance between adjacent pixels.
use of invariant sites in vicarious calibration is The actual response function of the ground
becoming more popular with the changing IFOV is often not square; for example, for
design and demands of new instruments. Vicar- MODIS, it is twice as wide cross-track as in-track
ious calibration using pseudo-invariant sites has because of time integration during scanning. For
become increasingly accepted as a fundamental most whiskbroom scanners, such as AVHRR
post-launch calibration method to monitor long- and MODIS, the actual size of the ground
term performance of satellite reflective solar IFOV is a function of the scanning angle (see
sensors. There are several common desired char- Figures 1.7 and 1.8).
acteristics of an invariant sitedfor example, This topic is significant because level-1 radi-
temporal stability, spatial uniformity, little or no ance data or level-2 reflectance data should be
16 1. A SYSTEMATIC VIEW OF REMOTE SENSING

FIGURE 1.6 Illustrations of the point spread functions (PSF) in two- and three-dimensional spaces (Zhang, et al., 2006).
Reproduced with permission of IEEE

FIGURE 1.7 Sketch of pixel geometry for the AVHRR to illustrate the variations in pixel size in the alongscan direction
(Breaker 1990).

corrected for geometric distortions before calcu- decommutating errors and detector failure.
lating geophysical parameters in order to obtain Potential remnant artifacts include banding and
a truly absolute geophysical parameter. The striping. In the past, these effects were ignored
details are discussed in Chapter 2. or artificially removed using cosmetic algo-
rithms during radiometric preprocessing. For
example, dropped lines are usually filled with
1.4.3. Image Quality Enhancement
the values of the previous lines or the averages
Imperfections or image artifacts are continu- of the neighboring lines. Strips can be removed
ously caused by the instrument’s electronics, by using simple along-line convolution, high-
dead or dying detectors, and downlink errors. pass filtering, and forward and reverse principal
Known artifacts include the scan-correlated shift, component transformations.
memory effect, modulation transfer function, To assist human visual interpretation, various
and coherent noise. Dropped lines and in- image enhancement techniques have been incor-
operable detectors also exist as a result of porated in many remote sensing digital image
 1.4. DATA PROCESSING SYSTEM 17
FIGURE 1.8 MODIS IFOV depen-
dant on the view zenith angle

processing systems. These enhancement methods atmospheric and surface information, atmo-
can broadly be divided into spatial domain and spheric effects must be removed to estimate
frequency domain categories. In spatial domain land surface biogeophysical variables, particu-
techniques, we directly deal with the image larly from the reflective and thermal IR data,
pixels. Figure 1.9 illustrates the effects of a linear since microwave signals are not very sensitive
enhancement technique. The pixel values are to changes in atmospheric conditions.
manipulated to achieve the desired enhancement. Clouds in the atmosphere largely block
In frequency domain methods, the image is first Earth surface information and make most
transferred to the frequency domain. That is, the optical and thermal-IR imagery useless for
Fourier transform of the image is computed first, terrestrial applications. Various cloud and
all the enhancement operations are performed on shadow detection algorithms have been devel-
the Fourier transform of the image, and then the oped, and cloud mask is one of the high-level
inverse Fourier transform is performed to obtain atmospheric products. However, this is still
the resultant image. an active research area, and more effective
Note that the radiometric properties character- and reliable algorithms are needed. For km
izing environmental conditions are artificially coarse-resolution imagery (e.g., AVHRR and
altered by image enhancement methods. Most MODIS), there usually remain many cloudy
image enhancement techniques for assisting visual or mixed cloudy pixels after applying the cloud
interpretation should not be performed before mask. Various solutions have been used to
quantitatively estimating biophysical variables. address this issue. One solution relies on
temporal compositing techniques, converting
daily observations to weekly or monthly data
1.4.4. Atmospheric Correction
based on maximum vegetation index or other
Since the observed radiance recorded by criteria; other solutions include replacement
a spaceborne or airborne sensor contains both of these contaminated pixels using smoothing
18 1. A SYSTEMATIC VIEW OF REMOTE SENSING

algorithms. This topic will be discussed in Image fusion is not distinguished from image
Chapter 3. merging or image integration, which at the pixel
For optical imagery, both aerosol and water level may be in many different forms, for
vapor scatter and absorb the radiation reflected example:
from the surface. There are two approaches for
• Multi-temporal images from the same or
atmospheric correction. The first one assumes
multiple sensors for change detection (e.g.,
known atmospheric properties, usually the total
merge TM images acquired at different times);
amounts of aerosol and water vapor in the atmo-
• Multi-spatial images from the same or
spheric column, which may be estimated from
multiple sensors (e.g., merge ETMþ
other sensors and/or other sources. Many atmo-
panchromatic and multispectral images);
spheric radiative transfer codes (e.g., MODTRAN,
• Multiple images of different spectral regions
6S) can be used to calculate the quantities required
from the same or multiple sensors (e.g., merge
for atmospheric correction. The second one
SAR with optical imagery, or visible bands
relies only on the imagery itself without any
with thermal bands);
external information. If the atmospheric informa-
• Remote sensing images with ancillary data
tion can be accurately estimated from other
(e.g., topographic map).
sources, the first approach is preferable, but quite
often we are not that lucky in reality. This topic Chapter 3 will present some of these tech-
will be discussed in Chapter 5. niques in detail.
For thermal-IR imagery, if we can acquire When we examine high-level satellite prod-
atmospheric profile information (mainly temper- ucts, it is surprising to see that most products
ature and water vapor) from sounding data, are mainly generated from a single sensor. For
atmospheric correction is straightforward. example, the MODIS team produces albedo
The split-window approach based on two products mainly from MODIS data, which is
thermal-IR bands, when no such atmospheric also true for MISR, MERIS, etc. The same
profile information is available, is often used to product from different satellite sensors may
estimate land surface temperature without atmo- have different characteristics (e.g., spatial and
spheric correction. The details are available in temporal resolutions, accuracy). Instead of
Section 8.2.2. asking the user to pick the “best” product, we
can generate a blended/integrated product
1.4.5. Image Fusion and Product from multiple-sensor products. Chapter 22 is
devoted to addressing this topic in the example
Integration
of leaf area index (LAI).
There are many cases where we need to inte-
grate image data through image fusion tech-
niques. Definitions of image fusion in the 1.5. MAPPING CATEGORY
literature are very diverse. Image fusion can be VARIABLES
viewed as a process that produces a single image
from a set of input images. The fused image We are in general interested in two types of
should have more complete information and is land surface variables: category and quantitative.
more useful for estimating land surface vari- The category variables represent the types of
ables. It can improve both reliability by using objects on the land surfaces and are usually map-
redundant information and capability by using ped out through image classification. The purpose
complementary information, as illustrated in of image classification is to group together pixels
Figure 1.10. that have similar properties into a finite set of
 1.5. MAPPING CATEGORY VARIABLES 19

(a) Original image

(b) histogram

FIGURE 1.9 An example of linear enhancement: original image and its histogram (a) and (b); linearly enhanced image and
its histogram (c) and (d).
20 1. A SYSTEMATIC VIEW OF REMOTE SENSING

(c) Enhanced image

(d) New histogram

FIGURE 1.9 (Continued).

 1.5. MAPPING CATEGORY VARIABLES 21
2. Selection of features: Classification is executed
based on a series of features in the feature
space. It divides the feature space into several
classes based on a decision rule. Instead of
using the original bands, they are often
transformed into feature space to discriminate
between the classes. Examples of features
include various vegetation indexes, principal
components and those from the Tasseled Cap
transformation, and other spatial, temporal,
FIGURE 1.10 Illustration of image fusion.
and angular features. The subset of features is
selected to maximally distinguish different
classes. An example of a classified image is a land
classes.
cover map. Figure 1.11 is a global land cover map
3. Sampling of training data: Training is the
mapped from MODIS data. The key steps in the
process of defining the criteria by which
classification process are as follows:
these classes are recognized and is
1. Definition of classification system (scheme): performed with either a supervised or an
This depends on the objective and the unsupervised method. Supervised training
characteristics of the remote sensing data. The is closely controlled by the analyst, who
purpose of such a scheme is to provide selects pixels from each class based on high-
a framework for organizing and categorizing resolution imagery, ground truth data, or
the information that can be extracted from the maps, while unsupervised training is more
data. A number of classification schemes have computer-automated and enables the user to
been developed for mapping regional and specify some parameters that the computer
global land cover and land use maps. The uses to uncover statistical patterns that are
IGBP land cover classification system for inherent in the data but do not necessarily
global mapping using MODIS data is shown correspond to classes in the classification
in Figure 1.11. scheme.

FIGURE 1.11 Global land cover classification map from MODIS.

22 1. A SYSTEMATIC VIEW OF REMOTE SENSING

4. Classification: A parametric or nonparametric visual interpretation was the technique commonly

decision rule, which is often called a classifier, used for extracting land surface information.
is used to perform the actual sorting of pixels Statistical analysis later became a more common
into distinct class values. There are various method for quantitatively estimating land surface
classifiers, such as the parallelepiped information. As seen from the following chapters,
classifier, minimum distance classifier, various inversion techniques based on physically
maximum likelihood classifier, regression tree based surface radiation models have become
classifier, and support vector machine the subject of mainstream research (Liang 2007,
classifier. They are compared with the 2008). It is necessary to provide an overview of
training data so that an appropriate decision these techniques. Since many inversion algo-
rule is selected for classification. rithms are based on forward radiation modeling,
5. Accuracy assessment: The classified results let us start with that first.
should be checked and verified for their
accuracy and reliability. The training data are
1.6.1. Forward Radiation Modeling
usually divided into two parts, one for
training and the other for validation. In the This is the process that links the pixel values
evaluation of classification errors, of an image with surface characteristics through
a classification error matrix is typically mathematical models (Liang, 2004). We will
formed, which is sometimes called mainly present landscape generation, surface
a confusion matrix or contingency table. and atmosphere radiative transfer modeling,
and sensor models.
The details of image classification techniques
are not covered by this book, but the basic prin- 1.6.1.1. Scene Generation
ciples and progress can be found elsewhere
Scene generation is a quantitative description
(Jensen, 2004; Lu and Weng, 2007; Mather and
of our understanding of the landscape. Strahler
Magaly, 2010; Richards and Jia, 2005). Some
et al. (1986) identify two different scene models
typical techniques for mapping land-use types
in remote sensing: H- and L-resolution models.
will be discussed in Chapter 24. At most spatial
H-resolution models are applicable where the
resolutions, the majority of pixels are mixed. If
elements of the scene are larger than the pixel
a pixel is required not just to be labeled as one
size, and L-resolution models are applicable
of the cover types but to estimate the percentages
when the converse is true. H-resolution scenes
of the cover types, it would be more challenging.
can be generated using computer graphics tech-
How to estimate the fractional vegetation
niques. One of the most well-known techniques
coverage within one pixel will be discussed in
is called the L-system. L-resolution scenes can
Chapter 13.
be generated using mathematical models or
GIS (geographic information system) techniques.
1.6. ESTIMATING QUANTITATIVE
VARIABLES 1.6.1.2. Surface Radiation Modeling
Given landscape composition and its optical
To drive, calibrate, and validate the Earth properties, we can predict the radiation field.
process models and support various applications, Three types of models characterize the radiation
high-level products of quantitative variables are field of the scene, and they are commonly used
much more desirable. How to generate these in optical remote sensing: geometric optical models,
products is the main focus of this book. In the early turbid-medium radiative transfer models, and
stages of remote sensing technique development, computer simulation models.
 1.6. ESTIMATING QUANTITATIVE VARIABLES 23
In geometric optical models, canopy or soil is given optical properties, distributed randomly
assumed to consist of geometric protrusions in the scene and oriented in given directions. In
with prescribed shapes (e.g., cylinder, sphere, one-dimensional canopy models, canopy
cones, ellipsoid, spheroid), dimensions, and elements are assumed to be randomly distrib-
optical properties that are distributed on a back- uted, but three-dimensional RT models can take
ground surface in a defined manner (regularly or into account the structural information of the
randomly distributed). The total pixel value is landscape, as shown in Figure 1.13. The further
the weighted average of sunlit crown, sunlit development of geometric optical models has
ground, shadowed crown, and shadowed incorporated radiative transfer theory in calcu-
ground. Figure 1.12 illustrates a simulated lating the individual sunlit/shadow components;
canopy field with an ellipsoid shape and the the resulting models are often called hybrid models.
calculated sunlit and shadowed components. In computer simulation models, the arrange-
Turbid-medium radiative transfer models ment and orientation of scene elements are simu-
treat surface elements (leaf or soil particle) as lated on a computer, and the radiation properties
small absorbing and scattering particles with are determined based on the radiosity equations

FIGURE 1.12 Projection principles

of the geometric-optical model for the
canopy with an ellipsoid crown and
the simulated canopy field.
24 1. A SYSTEMATIC VIEW OF REMOTE SENSING

FIGURE 1.13 DART (Discrete Anisotropic Radiative Transfer)(Gastellu-Etchegorry, 2008)

and/or Monte Carlo ray tracing methods. Atmospheric gases, aerosols, and clouds scatter
Figure 1.14 compares a photo of grass field and and absorb the incoming solar radiation and the
the simulated field using the Botanical Plant reflected and/or emitted radiation from the
Modeling System (Lewis, 1999) based on the surface. As a result, the atmosphere greatly modu-
ray-tracing technique. lates the spectral dependence and spatial distribu-
tion of the surface radiation. The atmospheric
1.6.1.3. Atmospheric Radiative Transfer radiative transfer theory is quite mature, and
The radiation at the Earth’s surface is disturbed many computer software packages (e.g., MOD-
by the atmosphere before being captured by the TRAN, 6S) have been developed to enable us to
sensor in the atmosphere (airborne sensors) or calculate all necessary quantities, such as path
above the atmosphere (spaceborne sensors). radiance and transmittance.
 1.6. ESTIMATING QUANTITATIVE VARIABLES 25

FIGURE 1.14 A photo of an actual canopy (left) and the simulated canopy field (right). Courtesy of Dr. Mathias Disney at
University College London.

1.6.1.4. Sensor Modeling a root sum squared sense in that domain before
being converted to noise equivalent spectral
Since common detector materials do not radiance. The noise processes originating in the
respond across the entire optical spectrum, electronics include quantization noise, bit errors
most sensors have separate focal planes and (in recording or transmitting the data), and noise
noise mechanisms for each spectral region. The arising within the electrical components.
sensor model can describe the effects of an Besides the sensor spectral response function
imaging spectrometer on the spectral radiance and radiometric noise, the sensor model can
mean and covariance statistics of a land surface. also include spatial effects using PSF and MTF.
The input radiance statistics of every spectral
channel are modified by electronic gain, radio-
1.6.2. Inversion Methods
metric noise sources, and relative calibration
error to produce radiance signal statistics that All inversion algorithms for estimating quan-
represent the scene as imaged by the sensor. titative variables can be classified into three
The sensor model includes approximations groups: statistical methods, physical methods,
for the spectral response functions and radio- and hybrid methods. Statistical methods are
metric noise sources. The spectral response func- mainly based on a variety of vegetation indices.
tions of each instrument can be measured and Physical algorithms rely on inverting surface radi-
provided by the sensor manufacturers. Radio- ation models. A new trend is to combine statis-
metric noise processes are modeled by adding tical and physical methods, which is referred to
variance to the diagonal entries of the spectral as a hybrid algorithm in this book.
covariance matrices. Radiometric noise sources
come from the detector and electronics. Detector 1.6.2.1. Statistical Analysis and Machine
noise includes photon (shot) noise, thermal Learning Techniques
noise, and multiplexer/readout noise. Since Statistical models have proven to be very
detector parameters are often specified in terms useful in various remote sensing applications.
of electrons, the noise terms are summed in They are usually created using ground
26 1. A SYSTEMATIC VIEW OF REMOTE SENSING

measurements. Because it is very expensive to observation-error covariance matrix and B is

collect extensive ground measurements under background-error covariance matrix. The first
various conditions, the major weakness of term Jb is to force the optimal parameters as
models based on ground measurements is closely as possible to background fields, and
limited representation. An alternative solution the second term Jo is to adjust parameters so
is to simulate remotely sensed data using a phys- that model outputs will be as close to the obser-
ically based radiation model that may have been vations as possible. Specifying R and B depends
calibrated and validated by field measurements. on the relative accuracy of background informa-
The key modeling components have been dis- tion and remote sensing data products. In
cussed in Section 1.2.5.1. extreme cases, if the errors of the “first-guess”
Different statistical methods can be used to values are extremely large, the final estimates
relate inputs and outputs of the model simula- will be decided from the fitting of the observa-
tions. Besides the conventional multivariate tions and will be close to the “first-guess” values.
regression analysis, different machine learning The high computational demands of optimi-
methods and other advanced statistical analysis zation approaches have led to the use of more
techniques have been used, such as artificial simplified surface reflectance models, rather
neural network (ANN) methods (Sections 13.3.3 than forcing optimization algorithm efficiencies.
and 15.3), genetic algorithms (Section 13.3.5), and One of the general trends in optical remote
Bayesian networks (Section 13.3.4). sensing is to use simpler empirical or semi-
An important drawback of most machine empirical models (Section 7.1.3). Optimization
learning techniques (e.g., ANN) has been their algorithms are used to estimate the parameters
lack of explanation capability. It is increasingly in these simple models. These parameters are
apparent that, without some form of explanation then related to surface properties.
capability, the full potential of trained ANNs
may not be realized. Many studies have focused 1.6.2.3. Look-up Table Algorithms
on mechanisms, procedures, and algorithms Optimization algorithms are computationally
designed to insert knowledge into ANNs expensive and very slow, performing the inver-
(knowledge initialization), extract rules from sion process with a huge amount of remotely
trained ANNs (rule extraction), and utilize sensed data. The look-up table (LUT) approach
ANNs to refine existing rule bases (rule has been used extensively to speed up the inver-
refinement). sion process. It precomputes the model reflec-
tance for a large range of combinations of
1.6.2.2. Optimization Algorithms parameter values. In this manner, the most
Optimization algorithms estimate the param- computationally expensive aspect can be
eters (x) of the surface radiation model by mini- completed before the inversion is attempted,
mizing the cost function J defined as follows: and the problem is reduced to searching a LUT
for the modeled reflectance set that most resem-
1
JðxÞ ¼ ðx  xb ÞT B1 ðx  xb Þ bles the measured set.
2 In an ordinary LUT approach, the dimensions
1
þ ðHðxÞ  yÞT R1 ðHðxÞ  yÞ ¼ Jb þ Jo of the table must be large enough to achieve high
2 accuracy, which leads to much slower online
(1.1) searching. Moreover, many parameters must be
where y is the observation vector, xb is the back- fixed in the LUT method. To reduce the dimen-
ground field (or first guess), H is the surface sions of the LUTs for rapid table searching,
radiation model operator, and R is the empirical functions are used to fit the LUT
 1.7. PRODUCTION, ARCHIVING, AND DISTRIBUTION OF HIGH-LEVEL PRODUCTS 27
values so that a table searching procedure to satellite observations and vice versa; (3) an
becomes a simple calculation of the local func- objective function that combines model esti-
tions, or a simple linear regression is executed mates and observations along with any associ-
instead of table searching for each angular bin ated prior information and error structure; (4)
in the solar illumination and sensor viewing an optimization scheme that adjusts forward
geometry. model parameters or state variables to minimize
Section 11.3.6 presents the LUT method for the discrepancy between model estimates and
estimating LAI, and Section 6.3.3 presents its satellite observations; and (5) error matrices
application to estimation of incident solar that specify the uncertainty of the observations,
radiation. model, and any background information (these
are usually included in the objective function).
1.6.2.4. Data Assimilation Methods The details are given in Section 11.4.
The values of land surface variables, estimated
using the methods from the previous sections
from different sources, may not be physically 1.7. PRODUCTION, ARCHIVING,
consistent. Most techniques do not take advan- AND DISTRIBUTION OF HIGH-
tage of observations acquired at different times LEVEL PRODUCTS
and cannot handle observations with different
spatial resolutions together. In particular, these Satellite observations must be converted into
techniques estimate only variables that signifi- high-level biogeophysical products using the
cantly affect radiance received by the sensors. In inversion methods outlined above through
many cases, the estimation of some variables a production system. Given satellite observa-
not directly related to radiance is desirable. tions and the inversion algorithm, production
Given the ill-posed nature of remote sensing of high-level products is not straightforward
inversion (the number of unknowns is far greater because of the huge amount of data. Creation
than the number of observations) and vast expan- of data information systems is absolutely neces-
sion of the amount of observation data, data sary. For example, NASA’s principal Earth
fusion techniques, which simply register and Science information system is the Earth
combine datasets together from multiple sources, Observing System Data and Information System
may be one solution. These techniques will be (EOSDIS), which has been operational since
presented in Chapter 4. The data assimilation August 1994. EOSDIS acquires, processes,
(DA) method allows use of all available informa- archives, and distributes Earth Science data
tion within a time window to estimate various and information products created from satellite
unknowns of land surface models (Liang, 2007). data that arrive at a rate of more than four trillion
The information that can be incorporated includes bytes (4 terabytes) per day. More and more infor-
observational data, existing pertinent a priori mation systems are supported by high-perfor-
information, and, importantly, a dynamic model mance computing capabilities.
that describes the system of interest and encapsu- The various levels of data used by the EOSDIS
lates theoretical understanding. are defined below. For some instruments, there
A DA scheme commonly includes the will be no Level 1B product that is distinct
following components: (1) a forward dynamic from the Level 1A product. In these cases, the
model that describes the time evolution of state reference to Level 1B data can be assumed to
variables such as surface temperature, soil mois- refer to Level 1A data. Brief definitions follow:
ture, and carbon stocks; (2) an observation model Level 0: Reconstructed, unprocessed instru-
that relates the model estimates of state variables ment/payload data at full resolution; any and
28 1. A SYSTEMATIC VIEW OF REMOTE SENSING

all communications artifacts (e.g., synchroniza- select the most suitable product, or combination
tion frames, communications headers, duplicate of products, for their specific needs. As remote
data removed). In most cases these data are sensing observations are generally merged with
provided by EDOS to a Distributed Active other sources of information or assimilated
Archive Center (DAAC) as Production Data within process models, evaluation of product
Sets to produce higher level products. accuracy is required. Making quantified accu-
Level 1A: Reconstructed, unprocessed instru- racy information available to the user can ulti-
ment data at full resolution, time-referenced, and mately provide developers the necessary
annotated with ancillary information, including feedback for improving the products and can
radiometric and geometric calibration coeffi- possibly provide methods for their fusion to
cients, and georeferencing parameters (e.g., plat- construct a consistent long-term series of surface
form ephemeris, computed and appended but status.
not applied to the Level 0 data). Land product validation has to rely on
Level 1B: Level 1A data that have been pro- ground measurements, which may be time
cessed to sensor units (not all instruments will consuming and very expensive. Because of its
have a Level 1B equivalent). importance, such product validation must
Level 2: Derived geophysical variables at the involve the efforts of the entire community.
same resolution and location as the Level 1 Sharing the validation methodologies, instru-
source data. ments, measured data, and results pilots the
Level 3: Variables mapped on uniform space- way to success and progress. One of the critical
time grid scales, usually with some completeness issues is the mismatch between ground “point”
and consistency. measurements and kilometer-scale pixel values
Level 4: Model output or results from anal- over the heterogeneous landscape. Up-scaling
yses of lower level data (e.g., variables derived “point” measurements using high-resolution
from multiple measurements). remotely sensed data is the key to addressing
Archiving and distribution of the huge this issue; more details are discussed in many
amount of data and products are also chal- relevant chapters.
lenging. NASA Earth Science information is
archived at eight DAACs located across the
United States. The DAACs specialize by topic 1.9. REMOTE SENSING
area and make their data available to researchers APPLICATIONS
around the world.
More details on data, production, and distri- Remote sensing has generated comprehen-
bution systems are presented in Chapter 23. sive, near-real-time environmental data, infor-
mation, and analyses. It serves a wide range of
users and empowers decision makers to respond
1.8. PRODUCT VALIDATION more effectively to the many environmental
challenges facing modern civilization. Figure 1.15
One of the key components of information depicts the linkage and flow of information from
extraction is validation. Without a known accu- remote sensing observations and other in situ
racy, the product cannot be used reliably and, data to societal benefits. Data can be used for
therefore, has limited applicability. With various driving, calibrating, and validating models and
land products available, users need quantitative decision support tools. The last chapter of this
information on product uncertainties in order to book illustrates how different remote sensing
 1.9. REMOTE SENSING APPLICATIONS 29

FIGURE 1.15 Linking Earth observations with societal benefits (CENR/IWGEO, 2005)

data and products can be used for monitoring

• Understanding environmental factors
land cover and land use changes and assessing
affecting human health and well-being.
their environmental impacts.
• Improving management of energy resources.
GEOSS identified nine societal benefit areas in
• Understanding, assessing, predicting,
which there was recognition that clear societal
mitigating, and adapting to climate variability
benefits could be derived from a coordinated
and change.
global observation system. As illustrated in
• Improving water resource management
Figure 1.16, the nine societal benefit areas are:
through better understanding of the water cycle.
• Reducing loss of life and property from • Improving weather information, forecasting,
natural and human-induced disasters. and warning.

FIGURE 1.16 GEOSS and applica-

tions (GEO, 2005)
30 1. A SYSTEMATIC VIEW OF REMOTE SENSING

• proving the management and protection of generally a nonlinear ill-posed problem, and
terrestrial, coastal, and marine ecosystems. use of more spatial and temporal constraints
• Supporting sustainable agriculture and by incorporating a priori knowledge and inte-
combating desertification. grating multiple-source data deserves further
• Understanding, monitoring, and conserving research.
biodiversity. Although remote sensing data products
have been widely used, significant disconnects
Some of these societal benefit areas are
between remote sensing development and
themselves complex clusters of issues, with
applications continue to exist. Some products
many and varied stakeholders. In each area
developed by remote sensing scientists have
there are observational needs for many vari-
not been widely used, and many variables
ables, with requirements for their accuracy,
required by land process models and decision
spatial and temporal resolution, and speed of
support systems have not been generated.
delivery to the user. The societal benefit areas
The product accuracy and application require-
are at widely varying levels of maturity with
ments may not always be consistent. Successful
respect to establishing user needs, defining the
applications are not static but evolve as new
observation requirements, and implementing
sensor, data processing, and network technolo-
coordinated systems. For example, the Weather
gies emerge. Improved coupling of remote
area is very mature, while the Health area is
sensing science and applications would be
relatively immature in the context of Earth
most advantageous.
Observation.

References
1.10. CONCLUDING REMARKS Breaker, L. C. (1990). Estimating and removing sensor-
induced correlation from Advanced Very High Resolu-
tion Radiometer satellite data. Journal of Geophysical
Significant progress in engineering issues
Research, 95, 9701e9711.
has enabled remarkable improvements in plat- CENR/IWGEO (2005). Strategic plan for the U.S. integrated
form and sensor systems that can be seen in Earth observation system In council Committee on
many indicators, such as signal-to-noise ratios, Environment and natural Resources.
resolutions, pointing accuracies, geometric and Gastellu-Etchegorry, J. P. (2008). 3D modeling of satellite
spectral images, radiation budget and energy budget of
spectroradiometric integrity, and calibration.
urban landscapes. Meteorology and Atmospheric Physics,
The immense amounts of data available from 102, 187e207.
satellite observations present great challenges. GEO (2005). GEOSS 10-Year Implementation Plan Reference
Considerable investments have been put into Document. pp. 209.
developing physical models to understand Jensen, J. R. (2004). Introductory Digital Image Processing.
Prentice Hall.
surface radiation regimes, and some of these
Lewis, P. (1999). Three-dimensional plant modelling for
models have been incorporated into useful remote sensing simulation studies using the Botanical
algorithms for estimating land surface vari- Plant Modelling System. Agronomie, 19, 185e210.
ables from satellite observations; however, Liang, S. (2004). Quantitative Remote Sensing of Land Surfaces.
development of realistic and computationally New York: John Wiley & Sons.
Liang, S. (Ed.). (2008). Advances in Land Remote Sensing:
simplified surface radiation models mostly
System, Modeling, Inversion and Application. New York:
suitable for inversion of land surface variables Springer.
from satellite data is still urgently required. Liang, S. L. (2007). Recent developments in estimating
Inversion of land surface parameters is land surface biogeophysical variables from optical
 REFERENCES 31
remote sensing. Progress in Physical Geography, 31, Richards, J. A., & Jia, X. (2005). Remote sensing digital image
501e516. analysis: An introduction. Springer.
Lu, D., & Weng, Q. (2007). A survey of image classification Strahler, A. H., Woodcock, C. E., & Smith, J. A. (1986). On
methods and techniques for improving classification the nature of models in remote sensing. Remote Sensing of
performance. International Journal of Remote Sensing, 28, Environment, 20, 121e139.
823e870. Zhang, P., Li, J., Olson, E., Schmit, T. J., Li, J. L., & Menzel, W. P.
Mather, P. M., & Magaly, K. (2010). Computer Processing of (2006). Impact of point spread function on infrared radi-
Remotely Sensed Images: An Introduction. Hoboken, NJ: ances from geostationary satellites. IEEE Transactions on
John Wiley & Sons. Geoscience and Remote Sensing, 44, 2176e2183.